blob: e25420218604c5aac0db3e28d89d3e282033a168 (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
|
#!/bin/bash
#SBATCH --job-name=run_expts
#SBATCH --account=bfqt-delta-gpu
#SBATCH --partition=gpuH200x8
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=32
#SBATCH --gres=gpu:h200:8
#SBATCH --mem=400G
#SBATCH --time=48:00:00
#SBATCH --output=logs/run_expts_%j.out
#SBATCH --error=logs/run_expts_%j.err
# Run experiments with models loaded locally
# This job needs 8 GPUs:
# - 4 GPUs for 70B judge model
# - 2 GPUs for PersonalizedLLM (embedder, reranker, extractor, main LLM)
# - Reserve for headroom
set -e
cd /projects/bfqt/users/yurenh2/ml-projects/personalization-user-model
# Create logs and results directories
mkdir -p collaborativeagents/slurm/logs
mkdir -p collaborativeagents/results
echo "Starting experiments at $(date)"
echo "Job ID: $SLURM_JOB_ID"
echo "Node: $SLURMD_NODENAME"
echo "GPUs: $CUDA_VISIBLE_DEVICES"
# Activate environment
source /u/yurenh2/miniforge3/etc/profile.d/conda.sh
conda activate eval
# Check GPU availability
nvidia-smi
# Add project to path
export PYTHONPATH="${PWD}/src:${PWD}/collaborativeagents:${PYTHONPATH}"
# Run experiments
cd collaborativeagents/scripts
# Quick test first (2 profiles, 2 sessions)
echo "Running quick test..."
python run_experiments.py \
--methods rag_vector \
--datasets math-500 \
--n-profiles 2 \
--n-sessions 2 \
--profile-path ../data/complex_profiles_v2/profiles_100.jsonl \
--output-dir ../results/test_$(date +%Y%m%d_%H%M%S)
# Full run (uncomment when ready)
# echo "Running full experiments..."
# python run_experiments.py \
# --methods vanilla,all_memory,rag,rag_vector \
# --datasets math-500,gpqa,aime \
# --n-profiles 100 \
# --n-sessions 20 \
# --profile-path ../data/complex_profiles_v2/profiles_100.jsonl \
# --output-dir ../results/full_$(date +%Y%m%d_%H%M%S)
echo "Experiments completed at $(date)"
|
